started · updated
Google DeepMind uses AI to mathematically verify Signal encryption
Google DeepMind researchers have utilized a new agentic AI framework, AlphaProof Code, to perform formal verification on key security components of the Signal encryption protocol. This process involves using AI agents powered by Gemini to generate millions of lines of mathematical proofs to ensure software functions as intended.
The project, which was completed in four weeks, represents the first time the actual code of a widely used encryption protocol has been formally verified. While the AI did not uncover new vulnerabilities in the specific portions examined, it successfully identified known security flaws that had been previously patched in older versions of the software.
Researchers noted that this task would typically require several years of human expertise. Google DeepMind suggests that this technology could eventually be applied to other critical systems, such as medical devices and power grids, to make advanced security checks faster and more cost-effective, though human oversight remains necessary.
Entities
AlphaProof Code · Gemini · Google DeepMind · Signal · Swarat Chaudhuri